Using the Torso to Compensate for Non-Minimum Phase Behaviour in ZMP Bipedal Walking

Abstract

In Zero Moment Point (ZMP) bipedal walking, the conventional method is to use the cart-table model for generating the reference trajectory [1]. However, due to modeling and tracking errors and external disturbances, such as uneven terrain, the generated trajectorymust be adapted by a stabilizer that uses sensory inputs from force and torque sensors placed in the robot’s feet. The problem with the carttable model is that it is non-minimum phase which causes a significant, undesirable undershoot in the ZMP in order to cancel the effect of disturbances. In this paper, a novel scheme is proposed for ZMP feedback stabilization that utilizes the upper body to balance the humanoid robot. This method increases the performance and robustness of walking by reducing the undershoot and maintaining a desired bandwidth. The effectiveness of the proposed scheme is demonstrated using simulation and open problems are discussed.